Reproduce “VOiLA: Vectorized Online Planning with Learned Diffusion Models for POMDP Agents” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Marcus Hoerger, Rishikesh Joshi, Rahul Shome, Ian Manchester, Hanna Kurniawati
VOiLA learns task-independent transition and observation models from simulation, then uses them with belief tracking for online POMDP planning. It distills diffusion samplers into compact samplers, while a likelihood model supports belief updates. In simulated navigation, it achieved 98% success on a held-out layout, versus 0% for two reinforcement-learning baselines; VTS reached 74%. It succeeded in 10 of 10 physical runs in one room and task.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.